Abstract. Language is, to observation, primarily a physical, biological phenomenon rather than a semantic one. We are accustomed to being told that
language should not be regarded as a technology, yet the study of logicalization, the process by which natural languages evolve their connective vocabulary
from lexical items, looks like nothing so much as a series of technological innovations on a minute scale, that is, innovations by which pre-existing physical
sub-devices are exploited for the production of novel effects. Any attempt
to construct from even a detailed understanding of logicalization a way of
approaching larger questions of language origins must pay close attention to
the evolutionary changes in the scale at which these innovations are to be
looked for. At later stages they represent relatively small changes to a large
and complex structure; at the earliest stages ancestral innovations would have
represented relatively large-scale changes to small and comparatively simple
In a physical account of logicalization, meanings are treated as physical
rather than semantic types, in particular, types of neural effects accessible to
the processes of speech production. But the short-term evolutionary facts of
the case require that such types be regarded as species, that is, as unions of
populations temporally ordered by an engendering relation, and sharing with
organic species the features that (a) they are non-classical sets, and that (b)
every member of such a species has ancestors that are not members of the
species. Ultimately the phylum of effects that we are prepared to regard as
pre- or proto-linguistic effects finds its ancestry in a class of exploitations of
wholly non-linguistic effects. Arguably some gestural vestiges of these very
early exaptations persist in human practice.
1. The Origins of this Essay
The main illustration of this paper, the connection of language origins with
throwing, was and remains a methodological illustration rather than an hypothesis.
The subject of the illustration, pointing, arises mainly from an evening graduate
seminar on language run jointly by David Hamlyn and Peter Winch at Birkbeck
College in 1966. There we considered the view that language is learned through
pointing, and the difficulty for that view that pointing was already a complexly
linguistic act. By contrast, the content of the illustration arose naturally from
a biological representation of meaning introduced as a vehicle for explaining the
phenomenon of logicalization: the historical process by which languages evolve
their connective vocabulary from lexical vocabulary, mainly from that of physical
relationships. The gist of that representation is this: Let a meaning be a physical
type characterizable in the language of a neuro-functional theory of the relevant
structures of the brain; in particular, let it be a type of neural effect accessible
Key words and phrases. language, logicalization, evolution, technology.
to the structures controlling speech production. Then even as much as we know
about logicalization forces us to regard such a type as a species, that is, as a union
∪P of a set P of populations of effects, temporally ordered by an engendering
relation, earlier populations engendering later ones. Like the members of organic
species, every member of a meaning has ancestors that are not members of that
meaning. Now on such a representation what holds for species holds for genera
and higher-order taxonomic divisions. In particular the phylum of linguistic effects
has ancestral phyla that are non-linguistic effects–which is perhaps to say little
more than that all linguistic beings have non-linguistic ancestors. It is the “little
more” that gives interest to the application of a theoretical idiom conceived in
minutiae of linguistic change to the largest linguistic change of all. But as regards
the illustration, if pointing is already a richly linguistic act, what were its nonlinguistic forbears and how were its effects engendered by theirs?
In general, philosophy students, attracted to big questions rather than to little
ones, become restive under a regimen of detail. So the illustration, although (or
perhaps because) they invariably prefer to hear it as a speculation, redeems a
trimester of tedium, and gives them something sensible to write about. That seemed
to be, through a number of offerings of the logicalization course, the principal role
of “the story of pointing”. It was the more general methodological questions that it
raised that piqued my curiosity about language origins; it was a remark of Deacon’s
([2], 52) that convinced me, so to speak, that God was an Anglican. Deacon refers
. . . a number of other theories that view language origin as
secondary to some more specialized adaptation, e.g., lateralized tool
manufacture and use (Doreen Kimura), accurate stone throwing
(William Calvin), etc. (my emphasis)
Someone, I supposed, who actually knows something about the subject has come
to the conclusion that my illustration suggested. Perhaps the general theoretical
framework can find an application here.
2. Language and Technology
It is widely protested that language cannot be regarded as technology, because
language constitutes too large and too significant a portion of what we are to be
usefully thought of merely as an instrument. One might as well say that digestive enzymes are technology because we use them in our gastro-intestinal tract to
assimilate protein. To be sure, we can learn techniques of effective speaking, but
speaking itself is not (ontogenetically) acquired as a technique for doing something.
A feral child that has not developed language does not in general acquire it later,
as she might acquire techniques of donning clothes. There does seem, however, to
be some connection at an evolutionary level, between our ancestors’ having become
technological and their having become linguistic. Derek Bickerton:
. . . if any species is to behave like ours, it needs . . . the physical
capacity to manufacture a wide range of artifacts–something that
our remote ancestors already had in the form of prehensile hands
with a powerful grip. It may well be that, without some such
morphological advantage, even a species with language might not
succeed in radically restructuring its behavior. It may even be that
the complex reasoning processes we now deploy in language could
not have arisen had they not been derived in some fashion from
equally complex technological processes that we mastered first. .
. For our purposes, it is enough to note the interdependence of
language and technology in radical behavioral change. ([1], 173)
Bickerton goes on:
Since the power to conceive logically precedes the power to create,
we may therefore assume that a radical improvement in conceptual
power (such as might result from the development of protolanguage
into language) logically preceded a radical improvement in artifacts,
rather than vice versa. Certainly all the technological advances of
recorded history have had not one iota of effect on the structure of
language. Ibid
A deeper reason for avoiding the language of technology in an explanatory theory
of language is that the term technology is not well-defined. People who speak of
technology in connection with the phylogenetically early stages of language development among hominids often have in mind predative and agrarian technologies,
the development and use of weapons, missiles, the production and use of fire, the
subordination and selective breeding of other animals, the refinement of crop types,
agriculture and so on. Asked for a general account of what technology is, they give
no reply. Fair enough: their purposes do not require it. But since ours do, it will
be as well to try to say at least sufficiently much about it to explain why the topic
is being raised at all.
The fact of the matter is that we haven’t said what language is either, and
we might very well use enzymes as artificial aids to digestion, enzymes produced
by methods that everyone would agree were technological. (Come to that, we
might also use language, in ways that everyone agreed were technological, to open
a door, or start a car, or run a bath.) But in neither the case of language nor
the case of technology should we expect to be able to provide a defining set of
severally necessary and jointly sufficient conditions. In the case of language we
have given the reasons, and corresponding considerations apply to technology. In
introducing the notion of linguistic effects, we observed that the set of linguistic
effects is a non-classical set. Every linguistic effect has non-linguistic ancestors,
since every linguistic agent has clearly non-linguistic ancestors. Between those
and our later, clearly linguistic ancestors there must have been ancestors who were
linguistic to some unclear degree or other, and the effects of their vocal practices can
presumably be said also to be linguistic to some degree or other. There is no point
in supposing otherwise, even if the steps that led from the one to the other were
in important respects incremental. We can adopt the same liberal attitude to the
development of technology. Some effects are technological effects, that is, produced
by technological means, and others not. We can find clear enough examples to
illustrate the difference. But every technological effect is engendered by earlier
effects, and in the ancestral of that relation will be effects that are not technological.
In the ancestral of the species of technological effects will be species that antedate all
species of technological effects. Again, without repeating the case, it is reasonable
to speak of degrees of technologicality if only to accommodate the transition from
our pre-technological to our technological ancestors.
Purely for local purposes, let us suppose that we can identify language with some
set of linguistic effects, and technology with some set of technological effects. It is
natural to take the question about the nature of language or technology to require us
to distinguish between current or recent activities or capacities that are regarded
as linguistic or technological from those current or recent activities or capacities
that are not. If then we are to speak, diachronically, about the set of all linguistic
effects or the set of all technological effects, we will need special notation to remind
us that that is what we are talking about. I introduce the following convention.
Suppose that “label” names a particular set, S of co-special items, that is, members
of the same species. And suppose that these items are the product of evolution,
that is, that they were engendered by previous items, that these previous items
were engendered by still previous items, and so on, until, tracing the ancestry far
enough we find ancestors that, for whatever purpose, we do not regard as being
of the same species. Let S 0 be the closure of S under the engendered-by relation
for that species. That is, (a) S ⊆ S 0 and ∀m ∈ S 0 , if S 00 is the set of items
that engendered m, then S 00 ⊆ S 0 . Then we adopt “label ” as the name of the
portion of S 0 ⊇ S that takes our interest. By this convention, if we apply the label
“language” to a set of clearly linguistic effects, and we apply the label “technology”
to a set of clearly technological effects, then “language ” and “technology ” will
label portions of the corresponding closures of these sets of effects under their
respective engendered-by relations. In practice we needn’t quite regard these sets
as closed under the engendering relation, only as including sufficiently much of
the domain of the relation as to enable us to trace the evolutionary emergence of
language on the one hand and technology on the other. If we are to understand the
set of linguistic (technological) effects diachronically, then we will not imagine there
to have been a first linguistic (technological) effect, since the set of such effects is
non-classical. Membership in such a set, like membership in a biological species,
takes a value in the closed unit interval. For such a set we ought properly to abandon
the two-valued idiom of e ∈ S 0 and e0 ∈
/ S 0 in favour of the more general idiom of
0 ≤ ∈ (e0, S ) ≤ ∈ (e, S ) ≤ 1. The adoption of the language of non-classical sets
puts us at odds with the language of closure under engendered-by, since that closure
must take us to effects properly outside the sets that interest us, hence the use of
“label ” as the name of a portion of S 0 . The notation is not intended to overcome a
difficulty of mathematical imprecision–it is intentionally imprecise–but rather one
of physical imagination. It reminds us that the relationships we are trying to think
about are somewhere near the 0-end of the sets of technological and of linguistic
effects. The question that I want to raise here is not about the relationship between
language and technology as Bickerton understands that question, but about the
relationship between language and technology . To put the matter another way, if
we can label the earliest considered class of items pre-, the next earliest proso-, the
next proto- and so on as our knowledge of Greek enables us, the question is more
like: what is the connection between pre-technology and proso-language or between
proso-technology and proto-language? The difficulty is that the best set of such
prefixes that we can manage will not do justice to the fine-grainedness of their
co-evolution at the earliest stages.
Let us suppose that pre-technological hominids scavenged large game, but lacked
the means of killing it themselves. In eating flesh made available through accident,
they were exploiting incidental non-technological effects of natural occurrences.
(It doesn’t matter for the illustration that they might also, perhaps even mainly,
scavenged the victims of quicker and more adept hunting species.) We may imagine
them coming upon animals that had fallen off cliffs, or upon animals unto which
bits of cliffs had fallen. Then an early step toward a technological replacement
for this rather haphazard and unreliable arrangement would be to take steps to
increase the probabilities of such events, say by panicking herds and loosening
rocks. By gradual refinements, relatively fresh large game protein is made available
by means that are undeniably technological, albeit at a very primitive level of
technology. The point that such a story is intended to illustrate is that, in its earliest
stages, technological innovation imitates nature, reproducing the incidental effects
of natural events by more or less the same proximal means observed in nature, but at
relatively more convenient times and intervals. Subsequent refinements reduce the
amount of effort expended to the same effect and perhaps generalize the method,
enabling an expanded field of deployment, and freeing the technology from the
haphazard and varied particularities of its original application.
But technological interventions, besides having anticipated outcomes, like naturally occurring events, also have incidental effects. And since the now-refined, to
some extent ritualized technological intervention has a familiar practiced regularity,
so too do its outcomes have noticeably refined incidental characteristics. And these
too can be reliably reproduced for their own sake if there is some benefit to be had
from them. So technological advance engenders further technological advance, with
later generations requiring, in general, greater resolving power of discernment for
their discovery, and finer motor control in the refinements of their exploitation.
Into all of this we must add the effects of the transmission of the practice to new
generations of practitioners with variations due to initially misapprehended purpose
and modifications imposed by juvenile musculation and altered social arrangements.
3. Grammaticalization
Now linguistic change, and particular, the process of grammaticalization, bears
some of these marks of technological progress. Lexical uses have anticipated, and
exploited but also incidental effects. Grammaticalization is a process by which these
incidental effects are exploited, isolated and refined. Consider the grammaticalized
uses of various forms of the verb go as verb endings (Latin) and auxiliaries (English)
in future tense formations. Compare
3.1 I am going to visit Aunt Sally
as an explanation of motion toward Aunt Sally
3.2 I’m gönna be sick
in the absence of any motion and in the presence of other cues as a warning of
imminent upheaval. The second use, that of the progressive future, has evolved from
the first use, which reports motion and its end. But it would not have done so if the
first use did not incidentally give generally reliable grounds for the anticipation of
whatever was given as the end of motion. It is that incidentally created anticipation
that enables the successful refined use of the form in the noticeable absence of any
motion at all, and eventually even in the presence of motion. Even in an in-flight
conversation, I’m going to be married need not be taken as an explanation of the
current journey. Parallel remarks would apply to the prosodically distinguished,
logicalized use of the string
3.3 Martyn swims as well as Robyn
The discretely-valued, conjunctive X ϕ’s as well as Y has evolved from the persisting smoothly-valued use that reports comparable ϕ-ing ability. However, it would
not have done so but that almost all instances of the older use reliably ground the
anticipation that both of the comparands ϕ. The pattern is widely repeated:
(1) α-use of vocable-string S has a main effect, e and (among others) an unavoidable subsidiary effect, e0 .
(2) S is used in circumstances that nullify effect e, but not e0 .
(3) S acquires a prosodic or other variant β-use, S 0 which has effect e0 but not
e, even when circumstances would not nullify e.
The crudity of the use, admittedly a fine-grained crudity, is strikingly present
in one form or another virtually everywhere in language. The surprise of its ubiquity must be a common experience for anyone studying these matters for the first
time. As many people are taught by their early training to see in the world the
beauty of a divine creation, they were also taught by their literary training to revere their language almost as an holie thing. But where we have learned to see
elegant concision, the study of linguistic change shows us a Rube Goldberg world
of unlikely improvisations and recycled spare parts of long disused machinery and
antique devices cobbled together. It would be to decidedly odd effect that I asked
of a guest recently ensconced in my favourite armchair whether she is capable of
being comforted. Yet there is nothing odd in the effect of my asking whether she is
comfortable–and no insult in my asking whether the chair beneath her is comfortable.1 There suddenly is a murky cluttered landscape of ramshackle, makeshift and
forgotten purpose where before we thought we saw fixed order. The change of attitude is Galilean. The immutable crystalline spheres are become messy collocations
of stuff from somewhere else.
Now, rhapsodies aside, none of this implies that human language itself in its
present highly evolved state should be understood as a kind of technology; nevertheless, some central kinds of linguistic innovations look like nothing so much as
relatively small-scale technological innovations (and, considering the present state
of nano-technology, not, comparatively, so small-scale as all that.) In retrospect, it
seems no small feat to distill an expression of futurity from an expression of motion
and end. Yet linguistic interaction seems to have done it by the use of the language
of motion without accompanying motion, leaving only what had been incidentally
present in the earlier use as its product.
It is the similarity of these linguistic changes to small-scale technological innovation that prompts the question as to the connection between language and
technology . Do all linguistic effects have technological but pre-linguistic ancestors, that is, ancestral effect-types in the intersection of language and technology whose
memberships in the set, T of technological effect-types must be assigned a markedly
higher value in the unit interval than their memberships in the set L of linguistic effecttypes? Is there an effect-type e such that
∈ (e, T ∩ L) = ∈ (e, T ) × ∈ (e, L) ∈ (e, L)?
1Here is an exercise: write a long book demonstrating the truth of this paragraph and citing
only vocabulary and constructions from this essay as examples. Let one of the chapters concern
the use of English verbs of vision.
Any proposed answer must be hypothetical, but we can give one that will at least
illustrate the sort of candidate we have in mind, leaving aside questions of plausibility and means of confirmation. What would the marks of such an item be? In
the following partial list, we must remember that e is itself a non-classical set.
(1) We would expect events of type e to be the incidental effects of some common physical activity.
(2) We would expect these effects to have been effects in linguistic ancestors.
(3) We would expect them to have neural and perhaps larger physical component effects.
(4) We would expect them, in the earliest instances, to reproduce effects of
a type e0 that are also reliably produced naturally, that is, other than as
incidental outcomes of technological interventions.
(5) We would expect the susceptibility to e0 effects to be an inherited trait,
present in some form from early infancy.
(6) We would expect the physical intervention that produced incidental eeffects to engender a refined physical intervention type that produced descendent e-effects without the effects to which earlier ones were incidental.
(7) We would expect to find descendants of those e-effects in every descendent
language , and so, given the specifications of the search, in virtually every
current human language.
The qualification that the effects should reproduce naturally occurring (e0 ) effects
amounts to the supposition that some inherited biological propensity was available
for simple technological exploitation. The proposed candidate propensity is the
tendency to track motion, that is the tendency to detect sudden motion within the
visual field and to adjust focus, eye and head motion to maintain a visual fix on a
moving object both when it is continuously and also when it is only intermittently
visible. In the adopted language of effects, the reliable (e0 ) effect is the effect on a
subject of an object moving within the field of vision. The descendent effect is the
tracking effect of a missile moving within the field of vision. Simply, the tendency
to track moving objects in general implies a tendency to track thrown objects. It is
the tendency to track thrown objects that this illustration takes as technologically
exploitable, since throwing an object reliably produces the effect, and incidentally
directs attention to whatever the object was thrown at. So the technological effects
are the general effects of missile-throwing and the incidental effects are the effects
on those present, namely that they visually track the missile and thereby focus upon
its target, which for the sake of the example, we can assume is prey. There is at
this stage in the development a reliable means of drawing attention to any feature
whether prey or no, namely tossing a missile at it. And that new technology is
likely to be learned by children in the course of their learning missile technology
more generally. However, there is no need to waste missiles to draw attention
to a feature, since companions will tend visually to pick up the trajectory of a
missile even when they have only intermittent visual contact with it, it is sufficient
to go through the motions of throwing. Thus what was mere incidental effect of a
technology having one purpose has become the central purpose of a new descendent
technology incapable of achieving the main effect of its immediate ancestor. Freed
from the old end, the motions of throwing can be refined so as unambiguously,
and more efficiently to achieve the new, and to achieve it, should the situation
demand, more discreetly. For the purpose, the important elements of the motions
are the direction and terminal position of the arm and hand. Incidental features of
throwing, such as the extension of the forefinger, which had provided a useful spin
to the projectile and precisely controlled its release, might be retained for sighting
along, but a gesture would adequately serve the function that was greatly reduced
in the flamboyance and vigour of its execution retaining only the extension of the
arm in a particular direction and with its particular final position. The result is a
low-energy action which is in certain visual respects like throwing, but which will
not be taken for throwing, an action that will, however, retain effects that have been
engendered by effects of throwing. Specialized variations of the action would mimic
certain features of the ancestral action, for example, one which propels the hand
forward in a roughly parabolic trajectory, thereby adding the effect of anticipated
distance to that of anticipated direction. Another that combines an arm motion
having the character of throwing with forward movement of the thrower in the same
direction would have the effect of inducing companions to follow. And so on.
If throwing were the technological ancestor of pointing, if the early elements
of pointing were elements of disarmed or ineffectual instances of throwing, the
development of the former from the latter would have had some of the important
marks of grammaticalization, which places vocables of the language (such as have,
going, to, follows, from and so on)in environments (such as I am going to stand
here until I have demonstrated that P follows from Q) that render their lexical
loading inert, that is, environments in which they cannot have an earlier causal
significance, cannot have some neural effects that at an earlier stage they could
not have been used without. But linguistic practices generally include a variety of
iterable devices for doing just that to any sufficiently small portion of the language,
as a temporary expedient for particular purposes. One such device is quotation,
in which strings of vocables, with undisturbed syntax, can be moved about safely
with none but a select and relatively passive set of reactive neural systems engaged
by their utterance. Mimicry is another. Here an ensemble of cues–some linguistic
possibly, some not–that are associated with one source is presented unmistakably
by another. And both quotation and mimicry can be iterated. We can mention the
mention of a string; we can mimic mimicry and so on. So strings and presentations
can be stripped even of the secondary neural effects left to them in first-order
quotation and mime. All such practices involve partial imitation. But once we have
the knack of noticing them, a limitless variety of such activities presents itself in
this light, among these, the creation of poetry. Consider Dylan Thomas’ They shall
be one with the man in the wind and the west moon in which the effect (whatever
it is) is achieved by transposition of the elements of the comparatively banal They
shall be one with the man in the moon and the west wind.
We have relatively easy investigative access to such relatively short term evolutionary linguistic developments as grammaticalization, though we do not yet have
a sufficiently general understanding of their neural dimension. But processes like
grammaticalization are processes that themselves have an evolutionary history. An
historico-linguistic study of classical Latin would perhaps show its perfect and future verb endings to be grammaticalized forms of earlier verbs of possession and
motion toward, and we might conclude that much the same exaptations as have
occurred in English occurred in Latin. But since, like us, the ancient Latins had
non-linguistic ancestors, there must have been much earlier stages in the ancestry
of Latin in which the elements available for grammaticalization were themselves
less evolved. After all, if the activities sufficiently early in the ancestry of language
were themselves non-linguistic, it follows that the grammatical distinctions of language themselves evolved. So the earliest ancestors of grammaticalizations were
themselves not grammaticalizations, but the products of some ancestral process to
which their ancestors were susceptible. The two millennia that separate us from
Cicero are insufficient to reveal much difference in the character of long-term linguistic descent. But it is a fair guess that a comparison with Indo-European or
ancient Hebrew would show more. We can therefore speak of grammaticalization ,
that is, the process of grammaticalization together with the catalogue of ancestral
processes that engendered it as a process of linguistic change. We cannot say with
any certainty that human languages of, say, 100,000 years ago had grammatical categories corresponding to those of any present-day language. But we can say with
assurance that however many categories they had, and whatever their character,
the items within those divisions had the uses they had because of the uses that
ancestral items had had within the divisions, whatever they had been, of ancestral
languages. It is tempting, perhaps to suppose that in the behaviours of sufficiently
early ancestors, there was but one undifferentiated grammatical category that could
not even be seen as such save by reference to descendent distinctions whatever they
were. But the illustration suggests a somewhat different picture: that at a sufficiently early stage, the grammatical, or rather grammatical , distinctions were
confused, in the best sense, with distinctions among actions more generally.
The idea is compatible with the notion of a relatively sudden development of distinctly linguistic complexity, and in fact, taken together with what we can readily
infer in broad terms about grammaticalization, would go some way toward explaining such a quick development were there independent evidence that one had
occurred. For even the development of species of pointing as grammaticalized , disarmed forms of throwing would create a suddenly enormous linguistic market for
a lexicon of nominal-like items beyond those already available through disarming
mimicry of distinguishing sounds. We need only suppose a growth in the resolution
of the distinctions represented within this early member of the genus lexicon to find
as well a slower accumulation of outworn vocables, those representing superseded
duller distinctions available for recycling as general verb-predicate-like items, that
is, available for a kind of grammaticalization that, though still generalized,is closer
to our normal understanding of the process.
Now this is far from presenting a hypothesis about changes in language being
brought about by changes in technology. Nor does it claim a relationship between grammaticalization and the development of syntax. But it does suggest a
common ancestry and early separation of the processes by which language and
technology evolved. And it can be understood as an illustration of a possible relationship between grammaticalization and syntax . If a hypothesis lurks here it
is a methodological one: deliberations about early evolutionary linguistic developments can benefit from a consideration of very early processes in the genus,
grammaticalization .
[1] Bickerton, Derek. Langusge and Species. University of Chicago Press. Chigago, 1990.
[2] Deacon, Terrence. Brain-Language Coevolution in [3], 49–83.
[3] Hawkins, John A. and Murray Gell-Mann editors. The Evolution of Human Languages: Proceedings of the Workshop on the Evolution of Human Languages, Held August, 1989 in Santa
Fe, New Mexico. Proceedings Volume XI, Santa Fe Institute Studies in the Sciences of Complexity. Addison-Wesley. Reading, Massachusetts, 1992.
—————————————————————Laboratory for Logic and Experimental Philosophy, Simon Fraser University, Burnaby BC, Canada V5A 1S6
E-mail address: jennings@sfu.ca